Rate limits

120 requests/minute per token (configurable per deployment via API_TOKEN_THROTTLE_RATE). The limit is counted per token, not per IP — session requests from the app are never throttled here.

Over the limit you get 429 with a Retry-After header and:

{ "code": "rate_limited", "message": "Rate limit exceeded. Please slow down.", "retry_after": 42 }

Back off for retry_after seconds and retry.

For batch jobs, spread requests or use one scoped token per integration. Agent sessions occasionally hit the limit during long multi-step tasks; they recover on their own.

Hitting 429 unexpectedly? See Troubleshooting API errors.